Making What AI Cannot

This project began with a simple question: what happens when emerging digital tools are pushed beyond visual ideation and used throughout the development of a real, technically demanding product?

A loudspeaker became the test case. AI was used for exploration and feedback, CAD translated ideas into manufacturable geometry, and FDM printing allowed concepts to be created, tested and revised. The finished speaker became both a functional product and a record of what these tools could, and could not, contribute to the design process.

What transferable knowledge emerges from a Programmatic Design Research process using Generative AI, CAD, and FDM 3D printing to design and fabricate a loudspeaker?

• Could an AI-assisted process produce genuinely useful design ideas?

• Could FDM printing become a manufacturing method rather than just prototyping?

• Could the resulting speaker actually perform properly?

AI Expanded the design space but it could not fully resolve an idea or translate technical requirements into a cohesive design rendering.

The generated concepts were not treated as finished designs. Ideas were assessed against proportion, manufacturability, acoustic requirements, and the overall character of the speaker. Useful elements were selected and recombined rather than directly reproduced

Physical testing was used to validate decisions and identify areas for improvement. Vibration analysis, acoustic measurement, and iterative prototyping helped to refine the enclosure, reduce resonance, and achieve balanced performance. 

This project showed that AI is most useful for expanding possibilities rather than making final design decisions. Its value depends heavily on the designer’s existing knowledge and ability to judge what is relevant, practical, and worth pursuing. At the same time, 3D printing proved capable of supporting more than rapid prototyping, becoming a viable manufacturing method when the product is designed around its strengths and limitations. Overall, the project reinforced that digital tools can accelerate exploration and iteration, but human judgement, testing, and hands-on making remain central to developing a resolved outcome.